[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Cellular Automata vs. Agent-Based Models
From: |
Marcus G. Daniels |
Subject: |
Re: Cellular Automata vs. Agent-Based Models |
Date: |
01 Oct 1999 09:51:41 -0700 |
User-agent: |
Gnus/5.070084 (Pterodactyl Gnus v0.84) Emacs/20.4 |
>>>>> "DT" == Dawn Trembath <address@hidden> writes:
DT> I'm now trying to understand the practical
DT> issue of why one would choose to use a CA representation, vs an
DT> ABM approach.
One thing Nelson didn't mention is that Swarm was designed to
facilitate fine-grained discrete event simulation. This is where the
actions that an agent performs have minimal mechanism and
iterative behaviors happen via the scheduling machinery.¹
Maximizing the information exposed to the scheduling machinery and
minimizing the amount of mechanism hidden in ad-hoc code brings an
agent-based model fairly close to CAs in terms of something you can
grab on to for statistical or analytical purposes. Essentially, an
exact time-indexed dataset with tags that reflect understandable
transformations of (typically local) state.
On the other hand, a basic feature of ABM is that you can engineer
agents to do things. You may actually know (or have a rich intuition
about) how some of the components in a model actually work. If you
want to use this kind of information for modeling, and this
information is complex or doesn't especially fit into a clean
theoretical framework, programming is a practical and precise way to
express it.
Although the traditional use for Swarm is to synthesize models, the
design also avoids obstructing the deconstruction of models.
¹ As the event distribution of a simulation gets more dense, having an
explicit representation of concurrent events also makes it possible to
amortize the weight of the extra scheduling abstraction (messages to
`Action' objects instead of pure compiled code) by distributing what
is logically concurrent to different physical CPUs.
==================================
Swarm-Modelling is for discussion of Simulation and Modelling techniques
esp. using Swarm. For list administration needs (esp. [un]subscribing),
please send a message to <address@hidden> with "help" in the
body of the message.
==================================